Machine-Oriented Compression

Dan Jacobellis

• Acknowledgments

• Abstract

• Introduction

  • Organization

• Machine Perceptual Quality

  • Background

  • Evaluation Framework for Machine Perceptual Quality

  • Key Findings

• Learned Compression for Compressed Learning

  • Introduction

  • Sandwiched Asymmetric Autoencoder via Wavelet Packet Transform

  • Implementation: WPT, Entropy Bottleneck, and Entropy Coding

  • Evaluation

• Lightweight, Versatile Codec Design

  • Introduction

  • Background and related work

  • Encoding-Efficient Asymmetric Autoencoder Design

  • Evaluation

  • Conclusion and Future Work

• Variable-Rate Compression and Projection-Pursuit Encoding

  • Introduction

  • Full-Input, Residual-Output Autoencoding

  • Evaluation

  • Conclusion

• Video and Real-Time Sensing

  • Introduction

  • Background

  • Related work

  • Local Inference with Delay-Conditioned Remote Assistance

  • Design and implementation

  • Evaluation

  • Bounded Performance Under Variable Delay

  • Conclusion

• Sensor-Embedded Autoencoding with One-Time Transcode

  • Introduction

  • Negative-Distortion Transcoding via Jointly Trained JPEG Proxy

  • Accuracy Gains from Negative-Distortion Transcoding

  • Conclusion